All day and then some for backup
For raw speed, it's 9.4 hours to back up all 8.5 terabytes. OTOH, it's 8.5TB!
Oracle has introduced the world's highest capacity tape drive, all raw 8.5TB of it, and the fastest too. The StorageTek T10000D transfers uncompressed data at 252MB/sec and 756MB/sec with 3:1 compression running. Its 252MB/sec data rate is 57.5 per cent faster than LTO-6's 160MB/sec, and LTO-6 only stores 2.5TB raw, less than …
For raw speed, it's 9.4 hours to back up all 8.5 terabytes. OTOH, it's 8.5TB!
No, all night. :)
Seems quite reasonable, for an off site backup started at 2100 it's finish before most people get in the next morning.
The real issue is recovery speed. That's a hell of a long time to wait to get all of the data back off, on the other hand I would imagine that recovering 8.5 terrabytes is going to be quite a protracted operation however you do it.
Also, 68 exabytes in one library? Why would you? You are not going to be able to restore most of that in any reasonable time, and if the room sets on fire you've just lost 68 exabytes in one go.
But my word, 8.5 terabytes on one drive in one night is pretty good going.
Sometimes you don't need fast restore. Think media libraries for content houses. You don't need ten years worth of accumulated stock footage, old textures, finished projects and various models immediately to hand, but artists draw on those resources often enough that you don't want to just delete them.
Don't be daft kids.
In the real world you wont have just one drive wired up to the robotics. You'll probably have 10 or so and you'll use parallelism to stripe the backups.
That's one hour **on paper** to do the restore once the thing has whirred away to read barcodes and fart around loading, seeking etc etc.
Note the lack of pricing information - no one outside a £100K+ pa IT budget need apply, you wont be able to afford the brochure let alone the sales commission.
"and if the room sets on fire you've just lost 68 exabytes in one go."
Ever heard of off-site storage? Apparently you haven't.
Who said the tape library would reside in the same data center?
Hey, I'm from the NSA you insensitive clod!
"if the room sets on fire you've just lost 68 exabytes in one go."
The point of a tape backup, is that you store the tapes in a different place to your servers, so you only loose 68EB in one go, not 136EB if you lost the backups and the live.
Also, why would you split up your backups into multiple chunks*, it'll just make your backups more complicated. ("Are the backups for marketing in tape room A or tape room B?").
* Obviously multiple full backups in different locations is a good idea.
Sounds like nice technology. Shame it's from Oracle....
I didn't spot the pricing for the media, but how does it compare to a couple of 4TB spinning platters?
Or, more to the point, how much would this solution cost compared to doing the same thing with LTO-6? Even if you need more tapes and drives, if Oracle go crazy with the pricing it might not be worth it.
The cost for the drive is well over $2000 ( eg http://www.backupworks.com/IBM-LTO-6-Drive-TS2260-3580S6E.aspx) without media, and for that price you can buy 50TB of hard drive. BUt someone has to pay for the El Reg journos.
When comparing tapes to HDD's you have to remember what your looking at.
HDD's are attractive for a single copy, however they have a low reader cost (ie; RDX reader) and then a freaking massive unit cost, because you have to buy expensive hard drives.
Tapes are attractive because whilst they have a more expensive reader, if your doing a 2 week rotation it usually works out cheaper because tapes cost bugger all.
I inherited an RDX setup where I work, which has a single week worth of cartridges. That's because the cost of simply moving to a second week would buy a small offices worth of new workstations. At previous jobs the cost of tapes for another week could be had out of petty cash without much fuss since the cashiers saw larger claims for mileage.
A 1TB RDX cartridge after formatting actually only holds about 900GB worth of data because it's a hard disc and they quote the unformatted size. Tapes actually hold what they claim to hold, which starts to matter quite a bit when you get to the edges of the quoted sizes.
Also when your dealing with disc to disc then it gets really expensive and so long term archive copies simply don't happen. Other places i've worked happily run a 2 week - 1 month rotation and then retire one tape a month to be a "archive". Like, everybody with tape does that to retire old tapes more than for the archives which are just nice to have. Do you know anybody doing that with RDX cartridges? I don't.
------ but then I realize where I heard all this before and why it all sounds familiar --
We have C's in our library -- the instant they show up we'll be putting in Ds.
As for restore time, so long as your backup isn't 100% single path linear you don't end up with all the data for one target restore on one tape, it usually (in our case) for REALLY large backups ends up spanning 4 or more tapes -- if you have the same number of tape drives on restore its actually quite skippy.
Grabbing one small file off one of those tapes however is something painfully slow.
And what all night? all weekend for some of these critters.
If you have to ask, you cannot afford it.
Really, tape can beat disk in per-gig costs, but the gap has closed - the break-even point is now so high, you need a truely ridiculous amount of storage before the economics of scale hand tape the price advantage. Something like this.
Quoting compressed capacity? Hah. Old trick.
- The 3:1 figure was almost true in the 90s. Back when all the bulky files were uncompressed and largely contained ascii text. Now days, no longer true: The big files are mostly compressed media, and even office documents have their own compression.
- The compression on tape rather sucks - even in hardware, you can't compress well while hitting the speed target. 3:1 is optimistic.
Suricou Raven said "The compression on tape rather sucks - even in hardware, you can't compress well while hitting the speed target"
In my experience, the above statement is ridiculous nonsense. Tape drives don't have any perceptible performance penalty in compression. Even doing compression+encryption in hardware on the drives I use the penalty is under 1%, and that's mostly the encryption (AES256).
By the way - to anyone looking at the data rate and thinking "that's not too fast" - compare it with a single spindle of HDD - it's a lot quicker.
You never use HDDs one at a time any more, and we [enterprise tape people] don't use tape drives like that either.
Someone was asking the cost of the media - it's usually cheaper than an HDD, but discounts vary massively. What's the life cost of keeping an HDD spinning versus a tape sitting in a slot using no power?
Anon because I work in this field, am well known, etc etc.
Your uninformed comment exemplifies the garbage spread by the "tape-is-dead" crowd. This baby delivers 8.5TB uncompressed.
Tape is certainly different to disk. If you use tape where disk is a better fit then it'll cost you in one way or another and strangely.... vice versa.
Sure. If you use the Office 2010 format.
I know quite a few big companies who have Office 2010. They are not going over all the documents they have ever created to convert them to the newest format.
So today, we still have loads of bulky ASCII files that are uncompressed. It's not because the newest version of something includes automatic compression that all companies set aside everything they do to spend weeks going over their archives and converting it all to the latest and greatest.
"compare it with a single spindle of HDD"
Compare tape to say 8 spindles in a RAID5 7:1 array that you might be using for such levels of storage, and tape is a lot slower even for sequential access. For random access speeds, tape is vastly slower.
"Sure. If you use the Office 2010 format."
Actually, Microsoft Office started using the "zip" archive format in 2003 for their Office Open XML .docx, .xlsx, .pptx, etc. files, which became the default file format with Microsoft Office 2007...
In modern tape backups you use the tape library as a host for the backups that you know you're going to need, but don't know when. All those tapes will be replicated either in an offsite vault, or robotised in a secondary datacentre.
DR recoveries come from either the disk snapshots you take for critical systems at an array/OS level or the disk backups in the primary backup tier. The library is used to host the backups that you need for either regulatory compliance or if you discover that your database corrupted a month ago and you need the version of the database from then. You also use these libraries for archive, or HSM where speed isn't really as much of an issue as capacity is.
Hmm, that should have been a reply up the thread somewhere. Too much beer and commenting about tape backup systems while drunk seems to suggest not enough life either...
Is there an affordable solution for backing up home computers on tape? I'm currently using blue ray discs, but it would be nice to have a tape that holds 0.5-1tb.
No, use a couple of external drives and treat them like tapes with a bundled drive (always have two on two different sites, never overwrite your current, etc....). Tape does not scale below some ten TB anymore.
For that size range look at LTO-3 to LTO-5. You might find an LTO-4 drive second hand for less than the cost of a 3TB hard disk. The ones I saw were all SAS, and a new SAS card will cost about the same as a pair of 3TB hard disks - perhaps you can find a second hand controller too. Then add the cost of the tapes. Tape is only economic and huge scales. Places that use them do upgrade when the price is right, and the kit is sufficiently reliable for a second hand market.
In the real world, you are not going to beat hard disks for backing up a few TB. If you really are burning 10-40 blue rays per backup, build one of these.
apart from the cost issue, with a single client you will usually not be able to keep the drive fed for streaming. Disks are much more forgiving, getting the minimum 40 MB/s for a LTO4 over several hours can be challenging.
I speak as someone who works in backup/data protection, currently as a software designer and formerly as an infrastructure designer for a couple of large UK FI.
Don't use tape at home. It's datacentre stuff. It's simply too expensive, a single tape drive just won't cut it, you've got to have two as a minimum, with one stored out of your house. There's no point in having tape backups, if when your house burns down, you end up with a stack of tapes and a bill for a replacement tape drive - if it's still made. You also need to keep the same hardware and firmware to make sure that you really can get that data back. I've seen too many tapes fail to play on drives which have different firmware levels.
Personally I use backup software that I'll not name as a backup to disk on a VM, all the VM is stored on a dedicated disk, not shared with anything else. Then, every Monday I bring my 3TB external disk from work home, stop the backup services and copy the backup data and catalog database backup to the external disk.
Of course, if you've got a load of data and a load of cash and can afford it, buy the current top of the range LTO drive and understand the risks but realise the advantages...
>>Is there an affordable solution for backing up home computers on tape? I'm currently using blue ray discs, but it would be nice to have a tape that holds 0.5-1tb.
Depends on volume, it's probable that a DIY NAS is a better solution, for example, I picked up a six bay/dual opteron server for £20 (stuck 16Gb in it for £40), then 5x2Tb drives for £280, nas4free gives 7.5Tb of RAID-Z protected storage sitting in my garage, with a whole bunch of services available (rsync etc.).
For backing this up I have another one (in sync, except for deletes which lag by 7 days).
But I also picked up an SDLT 320 / 15 slot auto changer for £50 (and put in a second drive I had spare), 15 tapes backs up around 3Tb, I only bothered using tapes because £50 was cheap and I got 200 unused tapes free that were otherwise being dumped, I do a full backup every couple of months (keeping two, one kept somewhere else) and weekly/nightly deltas.
My guess is that if you're managing with 25Gb blu-ray for backups you'd be much better off with disk, shop around and you'll find 3Tb USB for less than £80 (even at £1 for a 25Gb blank blu-ray, the same 3Tb would cost £120) or build a NAS, repurpose an old PC, stick it in your garage, you could even use it for torrents or whatever.
It should be 2.5TB native, not 3.2TB native.
Yup. The article suggests Oracle are positioning this against the IBM TS1140; a 3592 JC cartridge is 4TB native, but the LTO6 cartridges are indeed 2.5TB
"It might even be able to store all the hacks' notebooks used in El Reg's worldwide offices"
was that a hint to the manufacturers?
We had a client with DAT data tapes in a cartridge stack auto-feeder, that started the backups at Friday night, and had it finished halfway through Sunday. Along with regularly scheduled cleaning tapes at the right intervals, every six months the drive would fail, and they'd have it replaced under warranty.
You could say that it was driven past it's rated workload, but you could also say if you have a stacker that can hold 15 tapes, then why wouldn't you use 15 at a time? Hey, 30Gb was a lot in those days...
On another note, I see at least several suggestions on this comments list about using 4Tb drives as your backup. If you've spent some real time with SATA drives, you'll know that they fail. Unless you're using them with some class of RAID, you're going to lose data - and lots of it - 4Tb at a time. I have four Western Digital Black (desktop grade) drives, with one failure after 2.5 years - not that long. I have about 10 Western Digital Enterprise grade drives in the works, but they're still new - so I have no idea how they'll do (although quietly hoping they'll last longer than 2.5 bloody years). My only other experience is with enterprise grade SCSI drives, but they're a lot lesser capacity and probably not as suitable for long term storage - even if they virtually never fail. On the other hand, we have tapes that are so old, we no longer have the drives to read them (though storage recovery firms that can prove the tapes and data are still good).
Although tape is the preferred long term data storage medium, HDDs are still viable for home users who don't care for high reliablity (or more accurately, don't want to pay for it).
How a yank like myself ended up here is a mystery to me, but I work in this field and was cruising the T10000D announcement. Lots of good comments here, but I'll add a few:
* LTO 6 native cap. is only 2.5TB (not the 3.2TB quoted by article)
* The T10KD is priced extremely competitively to LTO ... don't look at list price, ask Oracle/STK for a quote. Customers often think they can't afford this drive and settle for LTO
* Oracle/STK sells LTO also, heck we make the LTO read/write heads at our fab - it has its place in the SMB market
* tape will always have a place in backup, you simply need an "air gap" for your data. You can snapshopt, replicate, mirror ... HDDs/disk all you want, but data corruption, application error, rogue employees, cyber attack ... can destroy/corrupt online/wired backups in a flash
* the media for this drive is the same that was used with the prior T10KC drive. So customers that bought the T10KC years ago can now get a 60% bump in capacity on their existing media!!! You will never see this on LTO, each new gen requires new media to get new benefits
* the big story for tape now is big data and archival data. Use a tiered storage methodology to move data to the right medium at the right time for lowest TCO ... Tier I - high performance disk, Tier II higher capcity disk, Tier III High Capacity Tape - Information Life Cycle Mgmt.
Why a tiered strategy? Maybe a 26 to 1 TCO might be compelling to a company that is interested in profitability - I posted with a link earlier not sure if that is allowed so search "Clipper Group Long-Term Storage TCo Tape Disk 2013" (the 2010 study showed a 15x TCO - so the gap between disk/tape is actually increasing as tape's ariel density increases at a faster rate)